Beskjeder

Dear all, I have updated the plans for the rest of the semester, see /studier/emner/matnat/fys/FYS5429/v23/timeplan/index.html?no-cache
There you will also find the video and whiteboard notes from the last lecture.
We end the course with a discussion of deep learning and generative models and unsupervised learning. This is a very active field of research. We start with Boltzmann machines as a stepping stone towards generative adversarial networks and variational autoencoders. 
Our last lecture will most likely be May 3, thereafter we focus only on project work and there are no lectures but we can always organized group or individual zoom sessions.

The deadline for the project is set to May 31, with some flexibility.  We use canvas for handing in.
The oral exam is something we can decide upon but I would like to suggest either June 14-16 or the week thereafter, starting with June 21.  The project counts 70% of th...

Dear all and welcome to FYS5429/9429.
Since we did not have a lecture last week, the plan for this week is to try to catch up again with what we started before the Easter break.
The plan is to cover generative models for the rest of the semester and we will discuss in more detail so-called Boltzmann machines this week.  This is our first step towards so-called  deep generative models and our reading recommendation is Goodfellow et al chapter 20.1-20.7, see https://www.deeplearningbook.org/contents/generative_models.html
Best wishes to you all,
Morten

Dear all, due to a meeting I have attend which conflicts with our schedule, I would like to suggest that we focus only on project work on Wednesday April 12 and that we have a regular lecture on the topic of  generative models and Boltzmann machines on April 19.
Also, I would like to suggest that we set a deadline for the project to May 31 and there is only one project. 
I am sorry for the late message but I hope this does not affect too much your plans for the week. 
It seems that everybody was comfortable with the project counting 70% of the final grade and the oral exam 30%. 
See you all next week and best wishes,
Morten

Dear all, I hope you are all doing well.
This week we will wrap up our discussion of autoencoders (AEs) and how these can be used in connection with the project. We will also link the discussion of AEs to the principal component analysis and finalize the discussion of the PCA method as well.  The lecture slides (missing some codes which will be added later today) are at https://github.com/CompPhysics/AdvancedMachineLearning/blob/main/doc/pub/week9/ipynb/week9.ipynb

Else, we will continue with lectures throughout the semester and will focus on unsupervised methods and so-called generative models. The plans (tentative) for the rest of the semester are  listed below here.
Best wishes to you all,
Morten

## March 20-24
- Autoencoders and discussions of codes and links with PCA
  - Reading recommendation: Goodfellow et al chapters 14
- Video of lecture at https://youtu.be/
- Handwritten notes at https://githu...

These two weeks we focus on autoencoders and how these can be used to study the project. We link the discussion of autoencoders with our discussion of the principal component analysis. We show various examples of different types of autoencoders.

Dear all, I hope you are doing well and that this week has started the best possible way. 
I would like to propose a small change of plans for this week, essentially since I am going to travel back to Norway on Wednesday.  I would therefore like to propose that we skip tomorrow's lecture (Wednesday March 1)  and rather try to organize a longer session on Wednesday the 8th from 1015am to 2pm approximately.
This will be in person and will hopefully allow us to plan better the project and discuss in more detail the mathematics of recurrent neural networks.
I am sorry for this last moment change, but I hope that by meeting next week on March 8 instead of tomorrow, we can get more out of the lecture. 
I hope this works for all of you. The session next week will also be recorded and we can meet at the CCSE at 1015am on March 8. In the meantime, I would like to recommend to study chapter 10 of Goodfellow et al at https://www.deeplearningbook.o...

The plans for this week are

• Discussion of the math and basics of CNNs
• Discussion of project 1
• Slides as t jupyter-notebook

• Discussion of project paths
• Convolutional neural networks (CNNs)

Slides at https://github.com/CompPhysics/AdvancedMachineLearning/blob/main/doc/pub/week3/ipynb/week4.ipynb 

Video of lecture at https://youtu.be/ku3sZ1s4G0E

Whiteboard notes at https://github.com/CompPhysics/AdvancedMachineLearning/blob/main/doc/HandwrittenNotes/NotesFeb152023.pdf

• Codes for feed forward neural networks (NNs)
• Discussion of first data sets to be studied

Slides at https://github.com/CompPhysics/AdvancedMachineLearning/blob/main/doc/pub/week3/ipynb/week3.ipynb 

Video of lecture at https://youtu.be/pB_WywRDnvg 

Handwritten notes at https://github.com/CompPhysics/AdvancedMachineLearning/blob/main/doc/HandwrittenNotes/NotesFeb82023.pdf

Dear all, first thx so much for having chosen FYS5429/9429. After our first meeting where we discussed possible research paths, I have made a questionnaire based on your inputs. Furthermore, I have also added possible options for the evaluation format. You will find the questionnaire at
https://docs.google.com/forms/d/18m0IY8_ElbtsEdTatxhO6JdSOGRHvWy5ZDe8oZoHL6A/edit

Also based on your inputs, we will have regular lectures which will also form the basis for a possible project 1. 
Unfortunately we cannot please everybody, however, the suggested lecture path overlaps well with those of you interested in advanced deep learning methods and physics driven ML.

We will start this week with a review of neural networks before we move over to convolutional neural networks and recurrent neural networks. The aim is to develop our own codes for this. That can also form the basis for project 1. After this we will discuss autoencoders and generative ad...

Welcome to a new semester and FYS5429/FYS9429.

This is a course based on self-studies but with weekly meetings which will also function as lectures and discussions of learning material.  The permanent zoom link is (also accessible from the weekly schedule)

FYS5429 zoom link
https://msu.zoom.us/j/6424997467?pwd=TEhTL0lmTmpGbHlnejZQa1pCdzRKdz09

Meeting ID: 642 499 7467
Passcode: FYS4411
 

Furthermore, all teaching material (it will be updated continuously) is available from the GitHub link at

https://github.com/CompPhysics/AdvancedMachineLearning

The recommended textbooks that we will follow are

o Goodfellow, Bengio and Courville, Deep Learning at https://www.deeplearningbook.org/

o Brunton and Kutz, Data driven Science and Engineering at https://www.cambridge.org/highereducation/books/data-driven-science-an...